staffing army rotc at colleges and universities · ... the other employs drilling reservists ......

Staffing Army ROTC at

Colleges and Universities

Alternatives forReducing the Use of

Active-Duty Soldiers

Arroyo Center

Charles A. GoldmanBruce R. Orvis

Michael G. MattockDorothy A. Smith

with

Rodger MadisonLaurie McDonald

Prepared for theUnited States Army

R

Approved for public release; distribution unlimited

The research described in this report was sponsored by the United States Army under Contract No. DASW01-96-C-0004.

RAND is a nonprofit institution that helps improve policy and decisionmaking through research and analysis. RAND® is a registered trademark. RAND’s publications do not necessarily reflect the opinions or policies of its research sponsors.

© Copyright 1999 RAND

All rights reserved. No part of this book may be reproduced in any form by any electronic or mechanical means (including photo-copying, recording, or information storage and retrieval) without permission in writing from RAND.

Library of Congress Cataloging-in-Publication Data

Staffing Army ROTC at colleges and universities : alternatives for reducing the use of active-duty soldiers / Charles A. Goldman . . . [et al.]. p. cm. “Prepared for the United States Army by RAND’s Arroyo Center.” “MR-992-A.” Includes bibliographical references. ISBN 0-8330-2644-5 1. United States. Army. Reserve Officers’ Training Corps. I. Goldman, Charles A., 1964- . II. United States. Army. III. Arroyo Center. U428.5.S73 1999 355.2 ' 232 ' 071173—dc21 98-35414

CIP

Published 1999 by RAND1700 Main Street, P.O. Box 2138, Santa Monica, CA 90407-2138

1333 H St., N.W., Washington, D.C. 20005-4707RAND URL: http://www.rand.org/

To order RAND documents or to obtain additional information, contact Distribution Services: Telephone: (310) 451-7002;

Fax: (310) 451-6915; Internet: [email protected]

iii

PREFACE

The increased tempo and range of military operations, coupled withreduced manning levels, are exerting pressure on the Army to use itsactive-duty soldiers optimally. Consequently, the Army is seekingopportunities to fill positions now occupied by active-duty soldierswith other personnel. Specifically, a recent Armywide Institutional/TDA Redesign Study called for the design and testing of staffing al-ternatives for the Senior Reserve Officer Training Corps (SROTC)program using a combination of Active Component, Reserve Com-ponent, or former military personnel. In support of this require-ment, RAND was asked to develop staffing alternatives and design atest of their effectiveness. This report discusses such alternatives anddescribes a test design to assess their feasibility for implementationthroughout SROTC.

In fiscal year 1997, RAND briefed the concept for the program andevaluation to the Vice Chief of Staff of the U.S. Army and senior offi-cials in the Army training and personnel communities. The ViceChief approved the concept and directed that the test begin in schoolyear 1997–1998. If successful, the program could also help to shapeand sustain the force in other areas, since it can be a prototype formaking similar substitutions for active-duty soldiers.

This research is sponsored by the Deputy Chief of Staff for Personneland the Commanding General, Cadet Command. It is being carriedout in the Manpower and Training Program of RAND’s Arroyo Cen-ter, a federally funded research and development center sponsoredby the United States Army. This research was conducted in 1996 and1997 and approved for public release in 1998.

v

CONTENTS

Preface ......................................... iii

Figures ......................................... vii

Tables.......................................... ix

Summary ....................................... xi

Acknowledgments................................. xv

Acronyms ....................................... xvii

Chapter OneINTRODUCTION .............................. 1Background and Purpose ........................ 1Approach .................................... 4How This Report Is Organized ..................... 4

Chapter TwoSTAFFING ALTERNATIVES ....................... 7Current Staffing................................ 7Staffing with Former Military and Other Civilians....... 8

Cost of Staffing with Former Military and OtherCivilians ................................. 11

Availability of Former Military and Other Civilians .... 14Staffing with Reservists and Civilians ................ 18

Cost of Staffing with Reservists and Civilians......... 21Availability of Reserve Personnel ................. 24

Summary of Cost and Availability Considerations ...... 26

vi Staffing Army ROTC at Colleges and Universities

Chapter ThreeEVALUATING THE SROTC STAFFING ALTERNATIVES .. 31Evaluation Design.............................. 32

Recruiting and Retention ....................... 32Workload................................... 37Advanced Camp Scores ........................ 39

Effect of School Year Cycle ....................... 40

Chapter FourCONCLUSIONS................................ 43

Appendix

A. ADDITIONAL DETAILS OF THE TEST PLAN .......... 45

B. MILITARY PERSONNEL COSTS .................... 51

C. GEOGRAPHIC AVAILABILITY MODEL............... 59

D. STATISTICAL POWER CALCULATIONS .............. 63

References ...................................... 67

vii

FIGURES

1. Typical SROTC Battalion Staffing................ 82. Former Military and Other Civilian Staffing

Alternative ................................ 93. Cost of Contracted Instructors.................. 124. Reservist and Civilian Staffing Alternative ......... 195. Workload Characteristics ..................... 216. Chance of Detecting Change in Recruiting

and Retention.............................. 347. Effect of a Second Year on Testing Change in

Recruiting and Retention ..................... 368. Chance of Detecting Change in Workload of

±5 Hours per Week .......................... 389. Chance of Detecting a 20-Point Change in

Advanced Camp Scores....................... 39

ix

TABLES

1. Replacement Qualifications: Former MilitaryAlternative ................................ 10

2. Cost Comparison: AC Versus Former MilitaryPersonnel ................................. 13

3. Replacement Qualifications: Reserve Alternative .... 224. Cost Comparison: AC Versus Reserve Personnel .... 235. Estimated Availability and Cost of Replacement

Personnel ................................. 296. Effect of Variation on Annual Commissions ........ 337. Evaluation Alternatives ....................... 41

A.1. Colleges and Universities Participating in the1997–1998 Staffing Experiment for ContractedFormer Military and Other Civilians.............. 46

A.2. Comparison of Experimental Units to OverallSROTC Program ............................ 47

B.1. Federal Benefits for Armed Forces Personnel ....... 52B.2. Annual Compensation Rates (1997$) ............. 57

xi

SUMMARY

This report analyzes alternatives to current Senior Reserve OfficerTraining Corps (SROTC) battalion staffing in which many active-dutysoldiers performing teaching or training functions would be replacedby other personnel: contracted civilians with prior military service orreservists. Additionally, civilians (with or without military experi-ence) would be contracted to help cover administrative and logisticsfunctions now performed by active-duty soldiers. We developed twoalternative staffing plans to be tested, each over a period of twoyears. One plan employs civilians with military experience workingfull time; the other employs drilling reservists working part time.

Feasibility and Cost of Alternatives

This analysis focuses on the battalion (school) level elements withinCadet Command. Of approximately 2,500 total Active Component(AC) authorizations for the command, 2,200 are for the battalions.Over time, it appears feasible to replace a significant number of theseAC soldiers by using either staffing alternative, perhaps at lower cost.If both options are found effective and implemented, the potentialsavings appear to be in the range of 700–900 assignments. The corre-sponding annual cost estimates for the replacement personnel are$45–$58 million (1997 dollars), not considering any savings fromeliminating AC positions. If only one option (former-military or re-servist) is implemented, the potential savings are about 100–200 as-signments lower than the 700–900 range. These savings estimates,however, are subject to considerable uncertainty that can be resolvedonly by actual experience with the program.

xii Staffing Army ROTC at Colleges and Universities

It appears that using reservists could save resources, if the personnelsavings are used to reduce AC endstrength. It is not clear whetherusing the former-military option would produce a net savings or costincrease; much of this uncertainty involves the cost of the fees thatthe Army will have to pay contractors to hire and manage replace-ment staff. Although it is likely to be less costly and might help covermore AC positions, the reservist option would take longer to phasein; it is not likely to be feasible on a broad basis for several years fol-lowing an implementation decision. Achieving potential savingsthus will take time. The Army’s actual cost experience will dependon how the options are phased in, on contractor fees, and on the mixof replacement staff among reservists, retirees, younger former mili-tary, and other civilians.

Assessment of Impact on SROTC Program

Although the staffing alternatives offer prospective benefits, they alsocarry potential risks. These include the possibility of adverse impactson the effectiveness and efficiency of the SROTC program in keyareas such as the number of officers commissioned, the quality ofcadet training, and cadre workload. A test should detect suchharmful impacts, which the Army might otherwise be forced to livewith on a long-term basis. To assure the Army of detecting evenlarge changes in these outcomes—for example, a 25 percent changein the number of commissions—our analysis indicates that a single-year evaluation would require assessing 25 trial battalions for eachstaffing alternative, a total of 50 units. Currently, there are 255SROTC host battalions. We considered smaller-scale trials, but theywould not have provided this crucial information with a reasonabledegree of confidence.

We also judged that a two-year evaluation period was preferable to aone-year trial: It would substantially reduce the number of battal-ions needing to be restaffed with trial replacements (15 versus 25 peralternative). The longer, smaller trial would thus reduce startup ef-fects in the test and cause less future personnel turbulence in theevent it is not successful and the positions need to be filled onceagain by AC soldiers. The two-year test would also have a somewhatlower cost—about $15 million in total by FY00 (1997 dollars), pri-marily for replacement staff costs. Last, because of the larger num-

Summary xiii

ber of units required in the single-year evaluation and the longer leadtime required to restaff these units, final results from the two-yeartrial of replacement staff at the smaller number of units would beavailable no later than those from the one-year trial: December 1999for the former-military option, and a year later for the reserve option(due to longer phase-in requirements).

The two-year trial began in the 1997–1998 school year for the former-military option. Three battalions began implementing options in-cluding reservists. The full test of the reserve option will begin in the1998–1999 school year.

xv

ACKNOWLEDGMENTS

We are grateful for the cooperation and counsel of the numerousArmy personnel who have contributed to this research. In particular,we wish to thank Cadet Command, the Office of the Chief of theArmy Reserve, and the Army’s Program Analysis and Evaluation Di-rectorate for providing data and sharing their analyses of the issuesinvolved in alternative staffing of SROTC. At RAND, the authors wishto thank Jerry Sollinger for helping to shape a great deal of raw mate-rial into a coherent report. Bob Bell contributed advice on statisticalmethods. Bob Bell and Glenn Gotz both made a number of helpfulsuggestions for improving an earlier draft of this report.

xvii

ACRONYMS

AC Active Component

ACPC Arroyo Center Policy Committee

AGR Active Guard/Reserve

ANCOC Advanced Noncommissioned Officer Course

APMS Assistant Professor of Military Science

ARPERSCOM Army Reserve Personnel Command

AT Annual Training

CPS Current Population Survey

DMDC Defense Manpower Data Center

G&A General and Administrative

GS General Schedule

IDT Inactive Duty Training

IRR Individual Ready Reserve

MOS Military Occupational Specialty

MUTA Multiple Unit Training Assembly

NCO Noncommissioned officer

OCAR Office of the Chief, Army Reserve

PCS Permanent Change of Station

xviii Staffing Army ROTC at Colleges and Universities

PERSCOM Personnel Command

PMS Professor of Military Science

PUMA Public Use Microdata Area

PUMS Public Use Microdata Sample

RC Reserve Component

ROTC Reserve Officer Training Corps

SIDPERS Standard Installation/Division Personnel System

SROTC Senior Reserve Officer Training Corps

TDA Table of Distribution and Allowances

TPU Troop Program Unit

TRADOC Training and Doctrine Command

USAR United States Army Reserve

UTA Unit Training Assembly

VCSA Vice Chief of Staff, U.S. Army

1

Chapter One

INTRODUCTION

BACKGROUND AND PURPOSE

The post–Cold War drawdown has exerted considerable pressure onthe Army to make the most effective use of its active-duty personnel.To do that, the Army is seeking opportunities to fill positions now oc-cupied by active-duty soldiers with other types of personnel, e.g.,civilians, retired military, or reservists. The Army commits a sub-stantial number of active-duty personnel to Senior Reserve OfficerTraining Corps (SROTC) battalions at colleges and universities.1

Army SROTC now faces significant pressure to reduce its resourceuse. Complicating this picture is the fact that program closures havenot kept pace with earlier staff reductions. As a result, the averagenumber of officers assigned per campus has fallen from 5.0 to 3.5over the past decade. Today, the reduced cadre are struggling to exe-cute their training mission while maintaining recruiting activity.With possible further reductions in endstrength looming, the Armyfaces a dilemma: It may have to further reduce the number of ActiveComponent (AC) personnel in the SROTC battalions, but it will bedifficult to close SROTC units. Such outright staff reductions jeopar-dize SROTC’s mission. If staffing substitutions can be achieved,however, Active Component commissioned officers and noncom-missioned officers could be reassigned to meet other force needs, or

______________1In addition to SROTC, there is also a Junior ROTC program (JROTC). JROTC uses veryfew active-duty personnel. Almost all JROTC instructor positions are filled by militaryretirees hired by local school districts.

2 Staffing Army ROTC at Colleges and Universities

the positions could be eliminated in compliance with drawdownrequirements. Thus, the Army is seeking new ways to staff SROTC.

Specifically, the Institutional/TDA Army Redesign Study’s UmbrellaIssue 41 raised the possibility of using a combination of Active Com-ponent, Reserve Component, and contracted retired officers andNCOs to manage and execute the SROTC program.2 In that context,this report identifies alternative staffing plans and evaluates them oncost and performance measures. The research includes two main el-ements: description of the options (staff mix, resource demands, andfeasibility) and a test design (evaluation requirements, site selection,and costs).

We will describe in detail two alternatives for staffing some of the in-structor and trainer positions in SROTC: One employs civilians withmilitary experience working full time, and the other employs drillingreservists working part time. These two alternatives do not reflectthe full range of options we considered. One option considered andrejected was the increased use of AGRs (Active Guard/Reserves, re-servists serving on full-time duty). We concluded that current legis-lation, practice, and cost would not permit wide expansion of the useof AGRs for SROTC duty.3 Another option considered and mostlyrejected was the use of civilians without former military experience.Only people with substantial relevant military experience are likely tohave accumulated the ability to perform the many functions requiredof instructors and trainers, including teaching military tactics,demonstrating customs and courtesies, recruiting, and acting as

______________2Generally, Army units that go to war are organized under a Table of Organization andEquipment and are referred to as TO&E units. Units that perform more administrativefunctions such as the Army’s schools are organized under a Table of Distribution andAllowances and are referred to as TDA units. The Institutional/TDA Army RedesignStudy catalogued a number of possible approaches for shifting some of the Army’sforce away from TDA functions and toward TO&E functions. Each of the possibleapproaches was grouped into an “Umbrella Issue.” Umbrella Issue 41 recommendeda test of staffing alternatives for SROTC battalions.3AGRs currently serve in SROTC battalions. Congress has authorized 275 AGRs to staffSROTC, although only about 206 are presently assigned. Because they tend to be moresenior in grade structure and years of service than AC soldiers, their pay andallowances are higher than those of equivalent AC soldiers. For these cost reasons andbecause Congress has authorized a very limited number of AGRs in total, we did notconsider it feasible to staff SROTC battalions with a significantly increased number ofAGRs.

Introduction 3

Army role models. We therefore concluded that teaching and train-ing positions should be reserved for people with requisite levels ofrecent military experience. For certain administrative and logisticalpositions, however, a wider spectrum of individuals appears feasible.In both options described, we open full-time administrative and lo-gistical positions to civilians with relevant military or nonmilitary ex-perience.

The Army could organize a troop program unit (TPU) for the part-time reservists assigned to each school, or some other administrativestructure might be used. Both of our staffing alternatives assumethat the Army will contract with a commercial firm that will, in turn,hire former military members and other civilians. An obvious optionis to have the Army or other federal government agency hire the per-sonnel directly. We rejected that approach for two reasons. The firstreason applies to retirees. Military retirees are required by law toforfeit all or part of their retirement pay if they accept a position withany organization of the federal government—a substantial disincen-tive. The second reason applies to all. If the Army rehired formermilitary members directly as civilians, the Army would face limitedcivilian personnel ceilings and would have to create a personnelsystem to recruit, hire, assign, evaluate, and develop these employ-ees. In view of the structure required and ceilings involved, the Armywould likely be better off retaining the personnel on active duty.

We began by considering all positions for replacement. However, weconcluded that some positions include duties that are inherentlygovernmental in nature and should be staffed with military person-nel. In addition, we considered it very important to retain a mini-mum of one full-time military member in each key functional area:instruction, recruiting, and training. Administrative and logisticalduties were not subject to this restriction; those positions currentlyfilled by military members may be reassigned to replacement per-sonnel. Overall, in the typical battalion this would result in twocommissioned officers and one noncommissioned officer as full-time military. The remaining active-duty positions—about four onaverage—could be filled with replacements.


APPROACH

Once we determined what type of personnel we could use and inwhich positions, three other tasks remained: determine the cost ofeach alternative, determine whether adequate personnel were avail-able, and design an evaluation plan.

We estimated the cost of present AC staff and potential replace-ments. Present AC staff and potential reservist replacements werecosted using Army pay and force structure information. To developcosts for former military members and other civilians, we combinedinformation from the U.S. Census with estimates of the costs of con-tracting.

We analyzed the potential availability of replacement personnel witha geography-based model that used U.S. Census data for former mili-tary personnel and other civilian replacements. For reservists, weused information from Army data systems in the same geography-based model.

We designed the evaluation plan to rely on existing measurementscollected as part of Cadet Command operations wherever possible.We defined three key measures to be used to judge the success orfailure of the replacement personnel: recruiting and retention(which measures future commission potential), cadre workload(which measures an important component of quality of life), and Ad-vanced Camp scores of cadets (which is a standardized measure ofcadet capability and training). These measures will be used to com-pare a selected experimental group of battalions with alternativestaffing against other battalions with traditional staffing.

HOW THIS REPORT IS ORGANIZED

The remainder of this report divides into three chapters and four ap-pendixes. Chapter Two discusses staffing alternatives, their impliedcosts, and their potential to replace AC soldiers. Chapter Three laysout an evaluation framework and draws relevant implications formeasures of effectiveness, the number of test sites required, and re-lated costs. Chapter Four presents our conclusions and the recom-mended test design. Appendix A describes the test implementation,reporting on the selection of schools and noting some differences

Introduction 5

from the plans presented in the main text. Appendixes B, C, and Dcontain additional details of the technical analyses for the computa-tion of military personnel costs, the geographic availability model,and the statistical power calculations.

7

Chapter Two

STAFFING ALTERNATIVES

This chapter describes how SROTC battalions are now staffed anddescribes two alternatives: one uses primarily civilians who are for-mer military personnel, and one uses primarily reservists. For eachalternative, we provide a cost estimate and compare it with the pres-ent cost of staffing the SROTC battalions with AC personnel. We alsoconsider the availability of the replacement personnel.

CURRENT STAFFING

In 1997, the typical SROTC battalion was authorized five officers,three NCOs, and one GS (general schedule) civilian as depicted inFigure 1. In addition, many schools funded a full-time or part-timesecretary to serve the SROTC department at no cost to the Army. Thenumber of assigned officer staff frequently falls below the authorizedlevel, as shown by the dotted box. Officer assignments at this timewere at most 80 percent of authorizations; NCO assignments wereabout 95 percent of authorizations, or higher. Because of differencesin program size and staffing availability, battalions’ actual staffingvaries both in terms of total number of personnel assigned and gradedistribution. The results presented in this report are based on the ac-tual staffing patterns in place in 1997. Although there have beensome subsequent changes in SROTC authorizations, the changeshave not yet affected the number of assigned personnel.

The four officers assigned are typically divided as follows: one lieu-tenant colonel, who is the Professor of Military Science (PMS) andcommander of the battalion; one major, who is often an AGR; andtwo AC captains. Aside from the PMS, all other officers are generally


RANDMR992-1

ADMIN/LOGGS

SECRETARYSchool provided

PMSAC O-5

TRNG NCOAC E-8

ADMIN/LOGAC E-5/E-6

APMSAGR O-4

APMSAC O-3

APMSAC O-3

TRNG NCOAC E-7

APMSAC O-3

PMS = Professor of Military ScienceAPMS = Assistant Professor of Military ScienceTRNG NCO = Training NCO

Figure 1—Typical SROTC Battalion Staffing

known as Assistant Professors of Military Science (APMS). There arethree noncommissioned officers: two are primarily responsible fororganizing and conducting training, and one is responsible for sup-ply and logistics, or, in some battalions, administrative duties.

These personnel perform tasks in instruction, training, recruiting,mentoring, administration, and logistics. The work occurs in threeprincipal settings: the college campus, field training, and summercamps. The discussion in this report focuses on the campus setting,although we also consider field training and summer camps.

STAFFING WITH FORMER MILITARY AND OTHERCIVILIANS

We considered a number of staffing options. After extensive discus-sions with members of the Training and Doctrine Command(TRADOC), the Office of the Chief of the Army Reserve (OCAR), andothers, we focused on two basic alternatives to the current staffing,which could be implemented singly or jointly. One is based on usingcivilians with former military service to replace instructors and train-

Staffing Alternatives 9

ers, shown in Figure 2. The figure compares a typical unit’s currentstaffing to the alternative staffing.

To expand the potential of this alternative to replace AC soldiers, weinclude both retirees and more junior individuals, provided theymeet specified conditions for length and recency of military service,as described in Table 1. The other staffing alternative, discussed laterin this chapter, is based on using drilling reservists on a part-timebasis.

In both staffing alternatives, we seek to replace as many AC soldiersas possible. (We are not concerned with replacing the existing AGRsoldiers assigned to SROTC.) Each alternative retains three full-timemilitary staff (AC or AGR), in the typical battalion, to ensure that full-time military staff can organize the key functions of management,teaching, recruiting, and training. Because inherently governmentalfunctions such as decisions on cadet progress and commissioningmust remain with government personnel, each alternative retains anAC or AGR PMS in every battalion. In the typical battalion, one singlefull-time soldier in addition to the PMS would not be sufficient tomanage the key functions, including recruiting. Therefore, in thetypical battalion, in addition to the PMS, each alternative retains oneAPMS and the senior NCO as full-time AC or AGR.

PMSAC O-5

RANDMR992-2

ADMIN/LOGGS


Current Staffing Alternative Staffing

ADMIN/LOGContracted

civilian

TRAINERContracted

former military

APMSContracted

former military

APMSContracted

former military

ADMIN/LOGGS


PMSAC O-5

TRNG NCOAC E-8

ADMIN/LOGAC E-5/E-6

APMSAGR O-4

APMSAC O-3

APMSAC O-3

TRNG NCOAC E-8

APMSAGR O-3

TRNG NCOAC E-7

Figure 2—Former Military and Other Civilian Staffing Alternative


Table 1

Replacement Qualifications: Former Military Alternative

Position Qualifications

Instructor Recent military experience(within 2 to 6 years)

Branch qualified O-3 or higherBachelor’s degreeLimit of 5 years in the job

Trainer Recent military experience(within 2 to 6 years)

E-7 or higherANCOC graduatePlatoon sergeant or branch-equivalent

leadership experienceLimit of 5 years in the job

Admin/log staff Military experience not requiredRelevant job experienceNo limit on tenure in job

Also, consistent with a plan being developed by TRADOC, our plansinclude contracted civilians to replace the NCO admin/log staff po-sition. Because of legal obstacles and given that the focus of Um-brella Issue 41 is on replacing AC soldiers, we do not contract out theadmin/log GS (civilian) position, although this could be done at alater date. Secretaries, if provided by the school, are also retained.

In the alternative that staffs the battalion using a combination of ACsoldiers and former military personnel, the model battalion has twofull-time officer instructors (one AC and one AGR), two former mili-tary members serving as APMSs, one full-time AC training NCO, andone former military member serving as a trainer. The enlistedadmin/log function is contracted to qualified civilians (with or with-out military experience).

Pursuant to our discussions with TRADOC, we established requiredqualifications for each of these positions. We require that instructorshave military experience equivalent to an O-3 or above (captain), be


branch qualified,1 and hold at least a bachelor’s degree. For thetrainers we require only significant military experience (equivalent tobeing in grade E-7). The contractor can fill the admin/log staff posi-tion with any civilian with relevant job experience.

Cost of Staffing with Former Military and Other Civilians

To estimate costs for the civilian replacements, we combine esti-mates of base wages, fringe benefits, and overhead. The base wagesare derived from the 1990 U.S. Census (5 percent Public Use MicroSample). The census data report years of active-duty military servicebut not the grade at separation. To match years of service to paygrade, we use the current AC force pay grade by years of service dis-tribution. Based on this analysis, we accept bachelor’s degree hold-ers with at least seven years of AC service as potential instructors(equivalent to O-3 and above).

We identify members in nonprofessional occupations with at least 16years of AC service as potential trainers (equivalent to E-7 andabove). For both categories, we restrict the population to those withmilitary experience within the past six years.

For instructors, wages are computed based on the census wages ofpostsecondary teachers with the cited military experience. For train-ers, nonprofessional job wages for persons with the cited military ex-perience are used. For admin/log wages, we use the wages of civilianclerical workers. In each case, the base wages were estimated by av-eraging over individuals who reported 48–52 weeks of work duringthe year and 30–50 average hours per week.

Figure 3 shows the results of our estimates for the instructors. Theestimates are calculated separately for nonretirees (7–19 years of ACservice) and retirees (20 or more years). Because retirees are moresenior, they have higher wages. Figure 3 displays the contribution ofthe three components—base salary, fringe benefits, and contractoroverhead—to the overall cost and compares that cost with that of anAC soldier.

______________1Branch qualification requires completion of the Officer Advanced Course and suc-cessful completion of company command or branch-equivalent command position.


��

��

RANDMR992-3

1997

dol

lars

(th

ousa

nds)

90

80

70

60

50

40

30

20

10

0Nonretirees JROTCRetirees

Contractoroverhead

Fringe

Basesalary

Present AC cost �

$20

$11

$35

$25

$14

$45

$13

$42

Figure 3—Cost of Contracted Instructors

To estimate the burdening and overhead that a contractor would ap-ply, we used TRADOC guidance on preparing an independent gov-ernment estimate for contracted services. We estimate fringe bene-fits at 31.6 percent. This figure is based on a sample of JROTC staffreceiving employee benefits such as health coverage and pensioncontributions, which average 11.6 percent of base wages, plus 20percent for legally required fringes such as FICA, disability insurance,unemployment insurance, and worker’s compensation.

We follow the TRADOC government estimate guidance to set over-head and G&A (general and administrative) expenses at a total of 30percent, and profit at 10 percent. These estimates are based on usinga national or regional contractor to recruit, hire, and manage staff.Such a procedure is likely to be used in the initial phase-in and testperiod of staffing alternatives. Upon full implementation, it might bepossible to use other, cheaper contracting forms. For that reason, wecompute a range for costs, assuming that fringe and overhead mightbe reduced by as much as one-third.

To assess the consistency of our wage estimates with known salariesof former military SROTC instructors, we compared the wages de-


rived from the census data to the wages of a sample of JROTC highschool instructors, all retirees. (This sample was collected for an-other RAND research project.) The U.S. Census wage estimates forretirees teaching at the college level are very close to the JROTCwages, as shown in Figure 3. JROTC instructors are employees of theschool districts, which pay at least half of the wages, plus all thefringes, and also cover any administrative overhead (not shown inthe figure).

Table 2 shows our estimate of the present cost of AC soldiers staffingSROTC battalions and the cost of their potential civilian replace-ments. Appendix B contains details of the cost model used to com-pute the costs of the present AC staff. The costs include base pay,allowances, permanent change of station (PCS) costs, retirementbenefits, and health care. In addition, the Army incurs costs formanaging and training the AC personnel assigned to Cadet Com-mand. Based on information from Cadet Command, we included anestimate of Cadet Command’s personnel management overhead (0.6percent of total costs). We did not include the personnel manage-ment overhead in other commands, such as PERSCOM.

These costs are based on the distribution of current instructors,trainers, and admin/log staff. In the typical battalion, we seek to re-place two instructors, one trainer, and one admin/log staff. (Sincethe actual staffing varies slightly from these averages, the figures inTable 2 do not add exactly.) The costs for instructors depend onwhether most positions are filled with retirees versus younger formermilitary members. With only retirees, costs are likely to exceed the

Table 2

Cost Comparison: AC Versus Former Military Personnel

Position (Number)

Present Costper AC Position

($000)

EstimatedReplacement Cost

($000)

Instructors (2) 76 56–84Trainers (1) 59 53–63Admin/log (1) 46 50–60Average total cost of replacement

staff per battalion 260 220–290


present cost of AC officers. With a mix, or with a less expensive con-tracting mechanism, costs are likely to be lower.

For trainers there is less of a range, since there is little opportunity touse less experienced former NCOs. Both trainer and admin/log staffcosts depend on the contract terms.

The total cost reported compares only the AC soldiers subject to re-placement to the cost of the replacements. If the replaced AC sol-diers are assigned to other duties, the cost of replacements repre-sents an additional cost to the Army of $220,000–$290,000 annuallyper SROTC battalion using alternative staffing. If the AC endstrengthis reduced, the Army may experience cost savings of up to $40,000($220,000 versus $260,000) or cost increases of up to $30,000($290,000 versus $260,000) annually for each SROTC battalion usingalternative staffing.

The cost of replacements is based primarily on the workload oncampus. There may be some additional costs incurred in arrangingfor former military personnel to attend field training exercises andsummer camps.2 Since the AC personnel are replaced on a one-for-one basis and replacements are costed on a 12-month contract, weexpect any additional cost to be modest.

Availability of Former Military and Other Civilians

Cost is not the only consideration. The likely availability of replace-ment personnel also has to be taken into account. We assessed thepotential availability of replacements for the AC SROTC battalionstaff members using available data on current staffing in individualSROTC units and, for the potential civilian replacements, 1990 U.S.Census data for the same geographic areas. We identified the localpools of qualified replacements according to the position qualifica-tions given earlier. We included all qualified persons within 50 milesof the local SROTC unit in the pools. Based on the replacementstrategy described above, we sought two replacement instructors,one replacement trainer, and one replacement admin/log clerk forthe typical unit.

______________2Contract vehicles often call for higher travel costs than are incurred for AC personnel.


For instructor positions, we further restricted the population to thosein professional, managerial, and skilled technical occupations whoreported earning less than the average base wages we expected thecontractor to offer. For instructors those wages correspond to thecalculations shown in Figure 3.

For trainer positions, we restricted the population to those in occu-pations other than managerial, technical, or skilled technical occu-pations who reported earning less than the average base wages weexpected the contractor to offer.

For admin/log positions, we did not require any military service; werequired only that individuals were in administrative or logistical oc-cupations and reported earning less than the average base wages weexpected the contractor to offer.

For the instructor and trainer positions, some adjustments were re-quired to scale the population of the 1990 census to reflect today’sformer military population numbers. Using the census data, weidentified records corresponding to qualified individuals with activemilitary service between September 1980 and the census date, April1990. We used the September 1980 cutoff since it was the most re-cent available date. Since the military has been drawn down sincethe 1980s, we adjusted the populations for instructor and trainerpositions. Enlisted numbers have been drawn down 32 percent (1.80million average during the 1980s versus 1.22 million at September 30,1996). Officer numbers have been drawn down 22 percent (299,000during the 1980s versus 232,000 at September 30, 1996). For instruc-tors, we reduced the population by 25 percent, slightly more than the22 percent drawdown.3 For trainers, we reduced the population by32 percent, the reduction in enlisted personnel.

Next, we converted the population estimates to annual rates of newmilitary members separating from the service. Since 9.5 years ofseparations were included in the September 1980–April 1990 period,we divided the population estimates by 9.5 to yield an annual esti-mated flow.

______________3The census data do not distinguish between officers and enlisted. We use a criterionof sufficient years of service with a bachelor’s degree. Since some (former) enlistedmay have a bachelor’s degree, we adjust the census population slightly downward.


Beyond these rather straightforward computations, there is substan-tial uncertainty over how many of the eligible population would beaware of the opportunity, qualified to do the work, and willing to ac-cept these jobs. We made a series of assumptions that we believe arebroadly reasonable, basing them on available data and judgment.But these assumptions have much uncertainty. To indicate the effectof variations up or down, we performed three different calculations:(1) a base case based on the factors described here, (2) a high esti-mate using double the population in the base case, and (3) a low es-timate using half the population in the base case.

We believe that most of those taking contracted positions would doso immediately after separating from service, during the first year. Inlater years, former military members seeking employment might alsoseek contracted SROTC positions. The census reports respondentswho indicated that they were looking for work. In the instructorpopulation, 8 percent of respondents were looking for work. In thetrainer population, the figure was 7 percent. In the admin/log popu-lation, it was 2.6 percent. These figures formed the base case.

We assumed that all of the military members who separated in themost recent year were seeking work. In each subsequent year weused the percentage looking for work. We analyzed two cases: six-year eligibility and two-year eligibility. Cadet Command recently hasexpressed a desire to limit eligibility to the more restrictive require-ment of two years since separation from service.

Since admin/log positions are not filled from a military population,we used the looking-for-work figure applied to the eligible popula-tion to estimate the number of people seeking jobs. There are noinitial-year and subsequent-year calculations for this group.

For each population, we then made several assumptions about whowould accept contracted employment. We assumed that 90 percentof the relevant population would be aware of the opportunity andthat at most 25 percent of those would be interested in Army SROTC.

For instructors and trainers we applied two additional factors: (1)the service individuals had served in while on active duty and (2) theeffect of a five-year cutoff on tenure in the contracted positions.These factors did not apply to the admin/log positions.


With respect to the first additional factor, we assumed that formermembers of services other than the Army would accept Army SROTCpositions at half the former Army service member rate (12.5 percent).

The second additional factor stems from Cadet Command’s imposi-tion of a five-year maximum term of employment for a contractedinstructor or trainer. We expect that a significant number of other-wise qualified and interested individuals will be discouraged by thislimitation. Using data from the Current Population Survey (CPS), weexamined the length of time on the job for populations comparableto our instructor and trainer groups. Because the CPS does not in-clude information on recency of military service, the available popu-lation breakdowns did not exactly correspond to our census groups.Based on our examination of the CPS data, we estimated that 50 per-cent of the instructor group and 60 percent of the trainer group couldexpect to be on a job longer than five years. We reduced the other-wise eligible and interested populations by these rates to account forthe effect of the five-year maximum term.

Because of the five-year cutoff, at least 20 percent of instructor andtrainer positions will have to be filled each year. We also expectsome departures before five years. Therefore we estimated that 33percent of the positions would require filling each year. Admin/logpositions are not subject to the five-year cutoff, so position turnoverwould likely be less than for instructors and trainers; we estimatedthat one-sixth of admin/log positions would turn over each year.

To estimate fill rates, we selected the eligible and available popula-tion for each position, using the factors described above. Wematched the population to each school’s average annual demand forpositions in the steady state, using the population within 50 miles ofthe school. The contractor could offer inducements for people tomove to where the jobs are, perhaps at higher cost to the Army. Wedid not include any such effects. Appendix C provides details of theprocedures used in the geographic availability model.

The results suggest that nearly all of the AC soldiers currently fillingadmin/log positions in SROTC units and the lion’s share of thosefilling targeted trainer positions could be replaced by qualified civil-ians. Replacement of AC instructors appears to be more problem-atic: It may be difficult to replace more than about 60 percent of the


desired number of AC SROTC instructors, even with the more gener-ous six-year eligibility. In total, the civilian (former-military) staffingalternative appears promising, covering about 720–760 of the 1,000positions currently filled with AC soldiers. Restricting instructor andtrainer positions to two-year eligibility yields estimates of about 720positions replaced. Allowing six-year eligibility increases positioncoverage by about 40 assignments.

More details of these estimates are provided in Table 5, at the end ofthis chapter. The table also reports availability estimates for the re-serve staffing alternative, which is described next.

STAFFING WITH RESERVISTS AND CIVILIANS

Turning now to the staffing alternative using Reserve Component(RC) personnel, we again retain two full-time AC/AGR officers andone AC NCO. We replace three full-time AC personnel with a groupof part-time reservists. Like the staffing alternative using formermilitary personnel, one of the current AC positions is filled by acivilian: the enlisted admin/log position. And like the former-military option, the GS admin/log and secretary positions are left un-changed. Figure 4 shows the current typical unit staffing and the al-ternative staffing for that typical unit.

A basic difference between the two staffing alternatives is that theformer-military alternative replaces each AC soldier with a single,full-time civilian, whereas the reserve alternative uses a group of re-servists working part time. The USAR states that each reservist willbe assigned five hours per week of duty during the academic year.The reservist will be paid for five drills each month based on thisschedule.4 Unlike traditional weekend drills, most of the reservistswill serve between Monday and Friday for part of a day or two. To

______________4Ordinarily a drill period lasts at least four hours. Five drills per month (e.g., a week-end drill running from Friday night through Sunday) represents at least 20 drill hours.The USAR intends to pay each reservist for five drills each month based on five hoursassigned per week times four weeks per month, which is also equal to 20 drill hours.Reservists on weekend drills often serve more than the minimum number of hours,and we expect the same in SROTC duty.


RANDMR992-4

ADMIN/LOGGS


Current Staffing Alternative Staffing

ADMIN/LOGContracted

civilian

TRAINER8 reservists

APMS8 reservists

APMS8 reservists

ADMIN/LOGGS


PMSAC O-5

PMSAC O-5

ADMIN/LOGAC E-5/E-6

APMSAGR O-4

APMSAC O-3

APMSAC O-3

TRNG NCOAC E-8

APMSAGR O-3

TRNG NCOAC E-8

TRNG NCOAC E-7

Figure 4—Reservist and Civilian Staffing Alternative

avoid overstating the number of replacements needed, the plan pre-sented is based on the conservative assumption that present staffwork about 40 hours per week, meaning that we will need eight re-servists working five hours per week to replace each of theofficer/NCO positions eliminated.5 Therefore, in the above battal-ion, to replace two APMSs and one trainer we need a total of 24 re-servists.

The USAR could organize these 24 reservists into a Troop ProgramUnit (TPU), within a larger administrative structure perhaps usingbrigades to group schools. The TPU members might be drawn frommembers of the Selected Reserves presently in other TPUs or frommembers of the Individual Ready Reserve (IRR), who would be con-verted to TPU status for SROTC. Alternatively, the USAR could acti-vate individual IRR members for SROTC duty without convertingthem to TPU status.

______________5If present staff work substantially more than 40 hours per week, then more than eightTPU members would be required to cover their duties. Because the actual hoursworked by both present staff and replacement staff are uncertain, we have designedthe test to use eight replacements.


To maximize efficiency in replacing the AC staff, we distinguishedthe duties performed by each of the replaced AC soldiers by a work-load analysis. Some rearrangement of duties may be required tomake the best use of reservists.

Although Cadet Command has done some internal studies of thetypical battalion workload, the studies are not comprehensiveenough to allow predictions of how workload would be affected bystaffing changes. We have used available information to estimate thedivision of duties into instruction, mentoring, recruiting, training,and other functions, which principally include administration andlogistics. These duties represent the ones performed during the aca-demic year on campus and at field training exercises. During thesummer, different patterns apply on campus and at summer camp.

Figure 5 represents the total workload for the four officers and fourNCOs in a typical battalion. One of the battalion NCOs is in a full-time admin/log position, as shown in the separated pie wedge. Asnoted, this position will be contracted out. Three of the remainingAC soldiers (two officers and one NCO) are to be replaced byTPU/IRR members. Based on our workload analysis, a reasonable,efficient replacement plan would consist of eight part-time instruc-tors, six part-time trainers, and ten part-time admin/log support per-sonnel. This plan shifts many of the mentoring and recruiting dutiesto the retained AC/AGR instructors. In return, some of their instruc-tion, training, administrative, and logistics duties shift to the part-time replacement staff. Provisions for field training exercises andsummer camps are discussed later.

In consultation with OCAR and Cadet Command, we specified quali-fications for each of the TPU or IRR replacements; Table 3 summa-rizes these qualifications. Instructors can be branch-qualified offi-cers at O-3 or above with a bachelor’s degree. Because senior NCOscurrently support instruction at some battalions, some of the part-time instructor positions could be expanded to include E-7 andabove, working in conjunction with officers. Trainers are primarilyE-7s with some E-6s and E-8s. The admin/log support replacementsare E-4s to E-6s in admin/log MOSs. Although E-7s could serve asadmin/log support, we limited replacements to grades E-4 to E-6 topreserve E-7s needed as trainers.


RANDMR992-5

Admin/log and other

23%

14%

16%

23%13%

11%

Training

Full-time admin/log

Instruction

Recruiting

Mentoring

Figure 5—Workload Characteristics

As in the previous staffing alternative, the admin/log position con-tractor has responsibility for finding members of the general civilianpopulation with the relevant experience for that position.

Cost of Staffing with Reservists and Civilians

We computed a cost estimate for the reserve replacements. The costestimates for the reservists are based on 48 annual drills, activetraining between 14 and 17 days (costed at 15.5 days), and 3 addi-tional days of active-duty training during the year for field trainingexercises or other full-time activities. We thus assumed that the av-erage reservist functioning in a TPU would serve 48 inactive-dutydrills and an average of 18.5 total days of active duty or annualtraining. We assume that if IRR members were activated for dutywithout conversion to the TPU, they would be paid an equivalentamount to the TPU soldiers. Appendix B contains details of the costmodel used.


Table 3

Replacement Qualifications: Reserve Alternative

Position Qualifications

Instructor(part-time)

Currently drilling; or recent active or reserve experience(within 2 to 6 years)

Branch qualified O-3 or higherBachelor’s degreeLimited time on the job

(to be determined)

Trainer(part-time)


E-7Limited time on the job

(to be determined)

Admin/log support(part-time)


E-4 to E-6Admin/log MOS

Admin/log staff Military experience not required(full-time) Relevant job experience

No limit on tenure in job

As shown in Table 4, we estimate that the reserve replacements areless expensive than AC soldiers. (For example, the cost estimate for areserve group replacing one instructor ranges from $66,000 to$73,000 compared to $76,000 for the AC position.) Overall, the use ofreservists to replace instructors and trainers and the use of con-tracted civilians to replace admin/log staff might save up to $30,000per battalion per year, compared to the full cost of AC personnel.These costs are based on the distribution of current AC instructors,trainers, and admin/log staff. In the typical battalion, we seek to re-place two instructors, one trainer, and one admin/log staff with eightreservist instructors, six reservist trainers, and ten reservistadmin/log staff, plus one full-time contracted civilian to performadmin/log duties. The relatively higher cost for civilian admin/logstaff slightly reduces the cost savings associated with using reservists.

Of the total costs for this option, about $50,000–$60,000 per battalionper year is the estimated cost of the contracted civilian admin/log


Table 4

Cost Comparison: AC Versus Reserve Personnel

Position (Number)

Present Costper AC Position

($000)

EstimatedReplacement Cost

($000)

Instructors (2) 76 66–73Trainers (1) 59 52–54Admin/log (1) 46 50–60Average total cost of replacement

staff per battalion 260 230–260

staff. About $10,000–$20,000 represents increased costs for the addi-tional days of reservists’ active duty or annual training over the nor-mal amount. The remainder of the cost is the present cost of TPUmembers who would be diverted from existing missions to SROTCduty. The reserve replacements are cheaper because on average thegrade level of the replacements is lower than that of the present ACstaff. In particular, the duties of the APMSs are being divided amongofficers and NCOs, where officers handle teaching responsibilitiesand NCOs handle some of the administrative and logistical supportresponsibilities that are currently handled by full-time officers.Moreover, as is true for the AC soldiers, we include the reservists’benefits and allowances in addition to their basic pay. For reservists,however, some of these entitlements are received only during thetwo-week annual training period, not during the 48 annual inactive-duty drills.

Although this analysis includes the costs of using reservists for on-campus and field-training duties, some costs associated with sum-mer camps are not included. This alternative staffing plan wouldrequire some reengineering in personnel assignments for summercamps. We have briefly examined the needs of the camps and be-lieve that the replacement plan would leave enough full-time AC andAGR personnel to staff key positions at camp and on campus over thesummer. Camps would have to use full-timers only in positions re-quiring continuity for the full summer, with reservists rotatingthroughout the summer to augment the full-timers. For example,the platoon staff positions might be reduced by one full-timer perplatoon. In place of one full-time position, reservists would rotate


during the relevant weeks. Some committees, such as physicaltraining, could be staffed only with reservists. Other committeeswould need a mix of full-time AC/AGR and reservists.

Reservists would come to camp for two weeks, with up to three extradays of duty authorized to smooth transitions. Depending on theposition, there would be four to six different reservists filling a singlerotating position. Travel costs to camp would increase, since severalindividuals would have to travel to camp each summer instead ofone. Lodging costs might also increase, if there is time overlap be-tween reservists.

Availability of Reserve Personnel

We now examine the potential availability of qualified reservist re-placements for the AC SROTC staff members in each battalion.Based on the replacement strategy described earlier, for the typicalunit we seek reserve replacements for two AC instructors and onetrainer, and one civilian replacement for the AC NCO admin/logclerk. Since members of the Selected Reserves and the IRR may beeligible to fill these positions, we examine both populations. Thepool of local reservists eligible for the positions was identified basedon SIDPERS data for USAR TPU members and ARPERSCOM’s IRRdatabases. For the civilian clerical replacements we again used 1990U.S. Census data for the schools’ geographic areas. We included allqualified persons within 50 miles of the local SROTC unit in thepools. As described in Appendix C, this analysis used a geographicavailability model similar to the one used for the contracted formermilitary replacement staff. Table 5 shows the AC replacements pos-sible when considering only members of the Selected Reserves nowin TPUs (the row “USAR TPU”). The table also shows the increase inreplacements when IRR members are used as well (the row “USARTPU + IRR”).

As in the case of the former military members, there is substantialuncertainty over the availability of qualified and motivated reservistswho could make time when needed for SROTC duty. As in the earlieranalysis, we estimate availability using three scenarios: a base case, ahigh case (double the base case), and a low case (half the base case).


In the estimates we make somewhat different assumptions about re-servists now serving in TPUs and those in the IRR. In the case of re-servists presently serving in TPUs whose primary duty would be in-struction, we used the pool of branch-qualified officers, O-3 andhigher, based on information from the Army databases. We assumethat all such qualified reservists who teach at the postsecondary levelin civilian life are qualified to teach, and that most teachers at thesecondary and elementary levels could qualify as well. However,some of these reservists may not remain in the Selected Reserves ifforced to accept SROTC duty, while others may not have time avail-able during the days and hours needed by the local SROTC battalion,especially during normal weekday working hours. These qualifica-tions and availability considerations lead us—based on educatedjudgment—to set the base case availability at 30 percent of otherwiseeligible TPU members who are postsecondary teachers and 15 per-cent of elementary and secondary teachers as being available to staffSROTC battalions. As in the earlier analysis, the high and low casesare based on double and half of these percentages, respectively.

In addition, some nonteacher TPU members with the specified mili-tary experience and bachelor’s degrees would be eligible. We as-sumed that about 60 percent would be qualified to teach, half ofthose would be interested, and one-quarter would have time avail-able. We therefore included 7.5 percent of TPU members other thanteachers with bachelor’s degrees.

Because we anticipate more difficulty in attracting IRR members—who are not currently committed to a TPU—for the base case, we in-cluded 5 percent of branch-qualified IRR (O-3 and higher) withbachelor’s degrees who had AC or drilling reserve service within thepast six years. We did not have occupation data on IRR members, sowe could not perform separate analyses for teachers versus non-teachers.

For training functions, the pool consisted of all local E-7s and 10 per-cent of E-6s and E-8s (to capture very senior or promotable E-7s aswell as junior E-8s). We assumed that all such persons were qualifiedand that 25 percent of this pool was interested and available. Foradmin/log support functions, OCAR defined a list of qualified MOSs.We assumed that 25 percent of this pool was available.


In this staffing option, as in the former-military option, full-timeadmin/log positions are filled by contracted civilians, who cover al-most all of the AC positions.

As is true for the earlier staffing alternative using former military per-sonnel, the results suggest that most of the AC soldiers currently fill-ing training positions in SROTC units could be replaced using the re-servist option. Replacement of AC instructors again appears to bemore problematic: In this case, it may be possible to replace abouthalf of the current AC SROTC instructors using TPU members and upto three-quarters if IRR members are used as well. In total, the re-servist staffing alternative appears promising, covering about 850 ofthe 1,000 positions currently filled with AC soldiers.

SUMMARY OF COST AND AVAILABILITY CONSIDERATIONS

Table 5 summarizes the overall implications of the availability andcost analyses for the two staffing alternatives, implemented bothsingly and jointly. Each row shows the number of assigned AC staffthat is replaced. The chart also shows the estimated annual cost (inmillions of 1997 dollars) of compensating the replacement person-nel. Since the earlier estimates are in the form of ranges, we used themidpoints of the ranges to derive these replacement costs. The totalcosts are determined by the number of replacements and the costper replacement.

For each alternative and population, the table reports the base case,a low case, and a high case. The low case is based on half the popula-tion of the base case; the high case is based on double the populationof the base case.

For some positions, there is variation in the coverage depending onthe assumptions about availability (as seen by comparing the low,base, and high cases). For civilian admin/log replacements, avail-ability is excellent. Because there are so many qualified civiliansaround the country, doubling or halving the assumed rates makesalmost no difference in the coverage for this position (247 versus 254versus 255 for the three cases). There are a few schools located invery remote areas, meaning that about 20 positions are uncovered,even with the high assumptions.


The instructor and trainer positions show more variation over thethree cases. Availability for trainer positions is higher for both op-tions, so the impact of the alternative assumptions is low. For ex-ample, in the former-military (six-year eligible) alternative, movingfrom the low to high case increases coverage of trainers from 155 to197, an increase of 27 percent. The population in the high case isfour times larger than in the low case; this quadrupling of populationhas a quite modest effect. The pattern for trainers is similar acrossthe alternatives.

For instructor positions, the population is smaller compared to re-quirements. This is apparent in the greater differences between thelow, base, and high cases. Looking again at the former-military (six-year eligible) alternative, moving from the low to high case increasescoverage of instructors from 223 to 389, an increase of 74 percent. Inthe two-year eligible alternative, the population is smaller because ofthe more stringent restriction on recency of service. The smallerpopulation shows greater sensitivity: 365 versus 188, or a 94 percentincrease from the low to high case.

The reservist cases also have limited populations for instructors. Inaddition, each school must have a minimum number of available re-servists in order to replace one full-time AC position. If there aresome available, but not enough to make a reasonably full group, wedo not credit a replacement. The exact procedure is documented inAppendix C. Because of this requirement, cutting the reserve popu-lations has a potentially severe effect on coverage. The TPU-only al-ternative shows this clearly. The low case fills 79 instructor positions,compared to 184 for the base case. In other words, reducing thepopulation to 50 percent reduces coverage to 42 percent. This occursbecause many schools fall below the threshold where they haveenough reservists to fill a single instructor position.

If the present IRR members are included, reserve populations in-crease and coverage improves. In addition, the larger population isless sensitive to the assumptions.

As the population of eligible individuals broadens, the coverage ratesbecome less sensitive to specific assumptions. If both the reserveand former-military options are used wherever feasible throughout


the country, the coverage is highest—and least sensitive to assump-tions.

In the base case, if IRR members are included, the reservist coveragelevel compares favorably with the former-military alternative (745 forthe reserve versus 718–759 for the former military). The costs per ACsoldier replaced are also comparable, using the midpoints of theranges presented earlier. However, the extent to which all of the re-quired TPU and IRR members could be converted to SROTC duty isunknown, and our conversations with OCAR lead us to believe thatthe conversion process would require several years. Clearly, there isuncertainty about the availability of former military personnel, al-though the timetable to recruit the replacements and to resolve thisuncertainty appears to be faster for the former-military alternativethan for the reservist alternative. Thus, in essence, there is atimeline-versus-cost tradeoff between the two alternative staffingplans. Both plans benefit from the good projected availability ofcivilians to fill admin/log positions.

Using both former military personnel and reservists to staff SROTCbattalions increases coverage, compared to either option imple-mented alone. The combined options appear able to cover 700 to900 AC assignments, 100 to 200 more than either option singly. Theestimated cost of the combined staffing is about $45–58 million (1997dollars).

The ability of the options to complement each other indicates that infull implementation it would be most desirable to have both formermilitary and reservists available for staffing. We expect that theactual mix of the replacement staff would vary by location.6

______________6Indeed, the belief that a mix of reservist and former military replacement staff wouldbe used (as locally available) in any permanent version of the SROTC restaffing pro-gram led the VCSA to recommend testing of former-military and mixed staff options,rather than former- military and pure-reservist options.

Table 5

Estimated Availability and Cost of Replacement Personnel

Item Instructors Trainers Admin/Log TotalCost

(1997 $m)

Desired replacements 530 223 277 1,030

Replacement options and populations

Former military (six-year eligible) Low 223 155 247 635 38.2Base 319 186 254 759 47.1High 389 197 255 841 52.7

Former military (two-year eligible) Low 188 140 247 575 34.9Base 285 179 254 718 44.3High 365 195 255 815 50.9

USAR TPU Low 79 109 247 435 24.9Base 184 149 254 587 34.7High 284 178 255 717 43.2

USAR TPU + IRR Low 197 116 247 560 33.4Base 334 157 254 745 45.5High 403 185 255 843 51.8

USAR TPU + IRR + former military Low 303 175 247 725 44.8(six-year eligible) Base 407 201 254 862 54.0

High 453 206 255 914 57.6

NOTE: Contracted civilians fill admin/log positions in all options.

Staffing A

lternatives

29

31

Chapter Three

EVALUATING THE SROTC STAFFING ALTERNATIVES

The previous chapter described in detail two alternatives for staffingSROTC battalions. This chapter describes how the two alternativescould be evaluated. We identified three outcomes key to the effec-tiveness of SROTC battalions: recruiting and retention, which indi-cates the unit’s future potential to produce commissions; the work-load of the cadre members, which represents a proxy for their qualityof life; and the Advanced Camp scores achieved by unit members,which represents the capability of the cadets and the effectiveness oftheir training. Because the alternatives reduce the presence of ACsoldiers in the SROTC battalions and use persons who may differfrom the AC staff they are replacing in experience, age, total hoursavailable for work, and other characteristics, the new staffing planscould have effects on the functioning of the battalions. The use of re-servists on a part-time basis is another notable change from currentpractice. Each of these changes could be for the better or worse. Inboth cases, we are concerned about the performance of each battal-ion as a whole. Therefore we use the three outcome measurementsat the level of the battalion: recruiting and retention, workload, andAdvanced Camp scores.

Collecting additional information about the effects of the staffingalternatives may be desirable. To understand program changes inmore depth, attitude surveys or interviews could be used. Whiletracking contract costs is a given, detailed analyses of staffing andimplementation costs could be conducted to assess the determinantsof costs and how actual costs differ from the projections in thisanalysis.


EVALUATION DESIGN

The proposed evaluation design divides SROTC battalions into threegroups. The largest group is not affected by the proposed staffingchanges—the “comparison group.” There are two, smaller groups inwhich the staffing alternatives are phased in on a trial basis: Oneuses the former-military option to replace AC staff; the other uses thereserve option. Recall that both options use civilians to fill theadmin/log positions currently filled by active-duty soldiers.

The design allows us to identify postreplacement changes in out-comes in the trial replacement units and to compare these changesto the trends in the same outcomes in the comparison group over thesame period. Given the current budgetary environment, it is in-evitable that SROTC funding, staffing, or management practicescould change during the evaluation period. For this reason, amongothers, the use of comparison units is especially valuable.

The evaluation plan treats each host battalion—together with anyextension centers and cross-enrolled students—as a single unit in theevaluation. There are presently about 258 units (host battalions),although a few may be closed during the experiment period, reduc-ing the number of schools in the comparison group.

Recruiting and Retention

The most crucial measure of the effectiveness of a specific SROTCunit is the number of commissions produced over time. Large varia-tions—up or down—have a major impact on the effectiveness andefficiency of the SROTC program. If the number of commissions istoo low, the Army cannot meet its officer requirement or must ex-pand SROTC substantially to do so. If the number of commissions istoo high, then perhaps major economies in SROTC can be achieved.If these economies are not realized, then excessive commissioneesare slotted into the IRR.

For these reasons, it is desirable that a trial of the alternative staffingplans have an excellent chance of detecting major changes in thefuture commissioning potential of SROTC. The contemplated testdesigns do not last long enough to observe a single cohort of cadetsfrom entry through commission. Instead, the design of the test ob-

Evaluating the SROTC Staffing Alternatives 33

serves all cadets in the program over the course of each year. Newcadets are recruited into the freshman, sophomore, and juniorclasses. Existing cadets retain or drop out from one year to the next.The evaluation computes the number of new cadets recruited intoeach class as well as the percentage retained from each class over afull year.

Because of the importance of this measure, it is desirable to have atest with a 90 percent chance of detecting large changes in the num-ber of prospective commissions. This 90 percent level still leaves aone-in-ten chance that we would miss an important effect in the test.If achieving a 90 percent chance is not feasible, it is imperative thatthe probability of detecting large changes exceed 80 percent. Any-thing below 80 percent, or a one-in-five chance of missing importanteffects, gives us little confidence that the test will reveal the effects ofalternative staffing. If the trial does not detect such prospectivechanges, they will become the new reality when the staffing alterna-tives are broadly implemented throughout the SROTC program—with potentially damaging effect.

The Army must decide how much variation from the current com-missioning rates is acceptable on a long-term basis. As seen in Table6, if the requirement is for 3,800 commissions annually, a 5 percentvariation means that the actual number of commissions produced bythe staffing alternatives could be as low as (about) 3,600. A 15 per-cent variation means that annual commissions could fall short of the3,800 required by almost 600 officers. A 25 percent variation meansthat annual commissions could fall short of the 3,800 required by 950officers. The same size variations could occur in the positive direc-

Table 6

Effect of Variation on Annual Commissions

Undetected Variationfrom Requirement

Future Number of Commissions(versus 3,800 requirement)

–5% 3,610–15% 3,230–25% 2,850

NOTE: The same size variations are possible in the positivedirection (i.e., 3,900, 4,370, or 4,750 future commissionsversus 3,800 required).


tion as well. Clearly, these are substantial deviations from the tar-geted number of commissions. However, there is a tradeoff betweenthe Army’s willingness to accept substantially fewer (or greater)commissions on a long-term basis and the number of alternativelystaffed units that must participate in the trial replacement to preventsuch an outcome.

We would like to know the effects of a staffing alternative as soon aspossible after its initial implementation, ideally after a one-year trialreplacement. To estimate the number of commissions that a unitwould produce under an alternative staffing plan, we can analyze thehistorical patterns in progression through the program—for exam-ple, first to second year, on-campus recruit to Basic Camp to thirdyear, etc.—and commissioning. We can then test for changes inthese patterns and use the changes detected to predict future com-missions under alternative staffing.

Testing requires a large number of units because there is a great dealof variation in the rates at which cadets enter or drop out of SROTCacross schools and from year to year. Specifically, suppose that in

RANDMR992-6

Cha

nce

of d

etec

tion

(%)

100

80

60

40

20

05 2515 7565554535

Battalions in each alternative staffing group

Minimum acceptable chance of detection

±25%

±15%

±5%

Figure 6—Chance of Detecting Change in Recruiting and Retention


the alternatively staffed units, recruiting and retention worsens tothe extent of a 5 percent decrease in projected commissions com-pared to before the staffing change (corresponding to a full imple-mentation change to 3,610 versus 3,800 required). As Figure 6 shows,our statistical analyses of historical data indicate that even with avery large number of test units, there is little chance of detecting thischange. In contrast, the historical data imply that we could be confi-dent of detecting the recruiting and retention falloff that translatesinto a 15 percent decrease in commissions (meaning a full imple-mentation change to 3,230 versus 3,800), but only by testing a mini-mum of 75 units (for a trial replacement lasting one year). That is,only with about 75 test units do we exceed an 80 percent chance ofdetecting the change (shaded area). Fewer test units are needed if weare willing to accept the risk that we will only detect a change of 25percent or larger (corresponding to a full implementation change to2,850 commissions versus 3,800 required). Between 25 and 35 testunits are required in a one-year trial replacement to have an 80–90percent chance of detecting the 25 percent change.1

Clearly, we do not want to fail to detect a major change in recruitingand retention that we would be forced to live with after full imple-mentation of a staffing alternative. Thus, we want to be sure we testenough units to detect harmful changes. At the same time, we do notwant to test more units than necessary because, if a staffing alterna-tive is harmful, we want to be able to undo it quickly, cheaply, andwith as little personnel turbulence as possible. One approach toresolving this tradeoff is to extend the test period to two years. Thisallows additional data to be collected at each unit and reduces the

______________1In computing the possible effect of changes on the rate of commissions, we analyzedvarious measures based on historical data. The power calculations reported in Figures6 and 7 are based on changes in the rate of noncontracted MSIIs (sophomores) con-tracting as MSIIIs (juniors). This rate is crucial to future commissions and representsan area most likely to be affected by replacement personnel. We believe this rate pro-vides a reasonable representation of the complete set of transition measures we antic-ipate using in the actual analysis of the trial replacements. To generate the powercalculations, we conducted a Monte Carlo simulation using actual historical data.Each trial of the simulation selected a random group of battalions from the historicaldata to be the experimental group. Other battalions were assigned to the comparisongroup. (Because two options may be tested simultaneously, the comparison groupwas reduced by twice the number of test units. For example, a total of 250 units wouldbe divided into two 15-unit experimental groups and one 220-unit comparison group.)Appendix D provides more detail on the Monte Carlo simulation method.


number of units required to test each staffing alternative. Thisapproach is illustrated in Figure 7.

As shown, to have better than an 80 percent chance of detectingchanges in recruiting and retention equivalent to a 15 percentchange in projected commissions requires a minimum of about 40units in a two-year test, compared with 75 or more in the one-yearassessment. To detect a 25 percent change requires a minimum ofabout 15 units in a two-year test, compared with 25 or more testunits over a single year. As we will see next, a 25-unit single-year testor a 15-unit two-year test also would detect important changes incadre workload (quality of life) and cadet performance (capabilityand training effectiveness).

A two-year test has the additional benefit that it reduces the impactof any startup effects in integrating the replacement staff into theunits. Since the experiment calls for replacing about half the per-sonnel in a unit, we are concerned about disruptions. But this con-cern is tempered because the average turnover in a unit is one-third(based on average three-year assignments). A turnover of one-half is

RANDMR992-7

Cha

nce

of d

etec

tion

(%)

100

80

60

40

20

05 2515 7565554535



±25% (2 years)

±15% (1 year)

±15% (2 years)

±25% (1 year)

Figure 7—Effect of a Second Year on Testing Change in Recruitingand Retention


higher than normal but not completely unusual. Using fewer unitsover two years does increase the risk that the results will be affectedby one or two exceptionally good or bad units. Therefore, we wouldnot want to reduce the number of units participating below about10–15 no matter how long the test lasts.

A longer test might allow a better chance to detect harmful changesthat do not reach the 25 percent level. Extending the test to four ormore years—or adding additional units—is an option that the Armymight consider at the end of a two-year test, especially if the pro-jected effects do not pass the test of statistical significance but are inan undesirable direction. For example, if the two-year test generatesdata that indicate a projected 20 percent decline in recruiting andretention but lack statistical power to confirm this result, continuingor expanding the test would help to confirm whether the projectionwas cause for genuine concern.

Workload

Testing the change in cadre workload follows the same principles astesting changes in commissions. Our object is to determine if theremaining full-time military cadre (AC, AGR, or civilian) in the exper-imental units have to work harder than their counterparts in com-parison units. We also would like to know if alternative staffing is soefficient that it reduces workload for the remaining full-time military.

For workload, however, there is no existing data system to collectinformation. We have therefore designed a survey to collect infor-mation on the workload of cadre members in SROTC battalions. Wepilot tested the survey during the 1996–1997 school year and willimplement it during the staffing experiment. Because we did nothave historical data to use in estimating the power of this measure,we made assumptions about the underlying distribution of work-load, considering variation across schools, over time, and among in-dividuals.2 Since this was a theoretically derived distribution, it was

______________2Specifically, we assumed that each observation of a specific individual’s workloadwould have a standard deviation of about 10 hours per week. The computations werebased on the variance that resulted from seven administrations per year of a newworkload survey that we designed for this purpose. On average, each experimentalbattalion would have three full-time military cadre members responding to the survey


RANDMR992-8

Cha

nce

of d

etec

tion

(%)

100

80

60

40

20

05 2515 7565554535



2 years

1 year

Figure 8—Chance of Detecting Change in Workload of ±5 Hours per Week

convenient to use a standard power calculation for testing equality ofmeans between the experimental and comparison groups. Theresults appear in Figure 8. As shown, an increase (or decrease) inworkload of five hours per week per full-time cadre member (aboutone hour per day) requires 15–25 units to have at least an 80 percentchance of detection in a one-year test. Over two years, about 10–15units are needed. If these theoretical calculations are validated bythe actual data collected, we will likely need just one year of adminis-tration of the workload survey to collect sufficient information toassess changes in workload during the experiment.

The workload survey is also suitable for use by the replacement staff,contracted former military, and other civilians as well as reservists.Although their responses to the survey do not form part of this out-come measure, they may generate useful insights about the changing

_____________________________________________________________in addition to the replacement staff. The unit of analysis here was the averagereported workload over all full-time military respondents during the survey period. Asdetailed in Appendix D, our assumptions implied that the workload average in a one-year test had a standard deviation of 6.2 hours per week; in a two-year test it had astandard deviation of 5.2 hours per week.


RANDMR992-9

Cha

nce

of d

etec

tion

(%)

100

80

60

40

20

05 2515 7565554535



2 years

1 year

Figure 9—Chance of Detecting a 20-Point Change inAdvanced Camp Scores

distribution of workload within experimental units. This data alsomay help to establish specific causes of the effects noted on full-timemilitary cadre workload or recruiting and retention performance.

Advanced Camp Scores

Advanced Camp scores capture training effectiveness, capability,and leadership potential for SROTC cadets entering their senior yearof the program. A large change in program effectiveness could resultin a shift in average camp scores of 20 points, which would meanabout 15 percent of cadets moving from above average to below av-erage (or vice versa).3 As Figure 9 shows, historical data on campscores show that such a change requires 20–25 units to be detected in

______________3Advanced Camp is currently scored on a 1,000-point scale. Passing scores are 700and above. Cadets failing Advanced Camp often drop out before completion. Sincethey do not complete all the scoring events at camp, they do not receive valid scores.Therefore, we focus on passing scores, which range from 700 to 1000. In 1995, themean passing score was 899, with a standard deviation of 83.


a one-year test. In a two-year test, 10–15 units are needed.4 Thus,again, the trial units required to detect large changes in commissionswould be sufficient to detect significant changes in Advanced Campscores.

EFFECT OF SCHOOL YEAR CYCLE

Thus far, we have discussed key measures of the effectiveness andefficiency of SROTC staffing alternatives and the number of trialunits required to learn how the staffing alternatives affect these out-comes. We now consider how the school year cycle affects the avail-ability of the required data.

The school year runs from about August to August. Workload infor-mation could be collected throughout the year, which would accountfor monthly variations in the types and amounts of work performed.Advanced Camp scores become available at the end of August orearly September. For the key outcome—prospective commissions—the required enrollment figures are not available until November.

The implication of this school cycle is that although we would getsome indication of the success of a staffing alternative by summerfollowing the school year in which it is implemented, it would be atleast December of that year before the full information on the suc-cess or failure of the trial in that year could be acquired and ana-lyzed. For this reason, replacement personnel must be retainedbeyond the end of the school year(s) tested. Given the time neededto make arrangements to release the replacements and restaff theSROTC units with AC soldiers in the event of an unsuccessful trial,the additional period during which the replacement staff need to beretained is on the order of one year following the trial.

The two options—former military and reservist—could be evaluatedover a one-year or two-year period, as shown in Table 7. Based onthe data shown in the previous charts, the recommended minimumsize for a pilot test would be 25 battalions per new staffing option fora one-year trial, or 15 for a two-year trial.

______________4To estimate these requirements, we used the same Monte Carlo method used in thepower calculations for commissions. See Appendix D.


Table 7

Evaluation Alternatives

Option Feature One-Year Trial Two-Year Trial

Former military and Battalions 25 15contracted civilians Duration Aug 98–Dec 99 Aug 97–Dec 99

Cost $12.7 m/2 years $11.4 m/3 years

USAR TPU/IRR and Battalions 25 15contracted civilians Duration Aug 99–Dec 00 Aug 98–Dec 00

Cost $3.6 m/2 years $3.2m/3 years600 TPU 360 TPU

Pursuant to our conversations with TRADOC and OCAR, we recog-nized that a two-year trial of the former military staffing option—which requires a smaller number of units—could be initiated in fall1997. In contrast, it would take until fall 1998 to contract the largerreplacement staff needed for a one-year trial. As a consequence, theresults of the two-year assessment would be available at the sametime as those of a one-year test. The timetables for the reservistalternative are analogous, but it would likely be fall 1998 or later be-fore the USAR could staff the 15 units required for a two-year test.

Using the costs developed earlier, we estimate the cost of the per-sonnel who would replace AC SROTC cadre as shown in the table.The estimates include only incremental costs for reservists; duringthe assessment period, we assume the test personnel are convertedfrom other duties within existing endstrength. We show the numberof reservists involved. We do not attempt to assess any impact onreadiness.

The estimates include the cost of retaining contracted replacementsan additional year after the test period while a decision is made onfull implementation of the staffing alternative throughout SROTC.Because these personnel must be retained in the alternatively staffedunits and because the two-year assessment requires fewer suchunits, it is less disruptive and somewhat cheaper than the one-yeartest, both in terms of annual and total cost (about $14 million forreplacement staff over three years and $1–2 million for assessment tocover both options). In addition, a two-year test will be less subject


to startup effects. These features and equivalent endpoint make atwo-year test preferable to a one-year test.

43

Chapter Four

CONCLUSIONS

This report has discussed alternatives to current SROTC battalionstaffing in which many active-duty soldiers performing teaching ortraining functions would be replaced by other personnel: contractedcivilians with former military service, or reservists. Additionally,civilians (with or without military experience) would be contracted tohelp cover administrative and logistics functions now performed byactive-duty soldiers. We recommend testing two alternative staffingplans. One plan focuses on former military personnel; the other fo-cuses on reservists.

Over time, it appears feasible to replace a significant number of ACpositions, perhaps at lower cost. But the final number of positionsthat could be replaced by either civilians (former military personnelfor teaching and training) or reservists must be determined. Com-bining reservists and civilians would further enhance coverage of ACpositions. The potential savings appear to be in the range of 700–900assignments. At the high end, this represents half of the AC soldiersin SROTC battalions. Achieving such savings would likely require thecombined use of reservists, former military personnel, and othercivilians (for admin/log positions).

It is possible that using replacement staff could save resources, if theAC staff savings are used to reduce endstrength. Although the use ofreservists appears slightly cheaper on average, it is not clear whethereither option would produce a net savings or cost increase. Thereservist–contracted civilian option would take longer to phase in,however, and is not likely to be feasible on a broad basis before FY00or later. Achieving potential savings thus will take time, and the


Army’s actual cost experience will depend on how the options arephased in, contractor fees, and the balance of replacement staffamong reservists, retirees, and younger former military personnel.

The evaluation design presented above will allow the Army to judgethe success or failure of the staffing changes in terms of remainingwithin broad ranges of the current levels of commissioning, cadetcapability, and cadre workload. A one-year test appears unwise,since getting information from a two-year trial of replacement staffwould disrupt fewer units, provide final results at the same time, andbe somewhat cheaper. Each option will need at least 15 units overtwo years in order to generate sufficient data for evaluation. We es-timate the costs of this test plan at about $15 million (1997 dollars)over three years, primarily for replacement staff. A test with fewerunits—or over a shorter time—will not provide crucial informationon the viability of the alternative staffing plans. In particular, it isunlikely to detect harmful changes in the SROTC program that theArmy would be forced to live with on a long-term basis.

45

Appendix A

ADDITIONAL DETAILS OF THE TEST PLAN

We presented an earlier version of this analysis to the VCSA in Jan-uary 1997. The VCSA approved a test based on the two-year test rec-ommendation. The contracted former-military and civilian optionwas implemented in 15 battalions during school year 1997–1998.Plans are now being made to implement the reserve option in schoolyear 1998–1999. Each set of battalions will be evaluated over the two-year period recommended above.

We and Cadet Command selected units to participate in the testbased on several considerations. Selection avoided units whose PMSwas scheduled to rotate during the summer of 1997 in order not toincrease the turbulence arising from the replacement of 3–4 otherpersonnel in the unit. The selection attempted to balance the unitsto match the distribution of all SROTC units in terms of geography,public versus private school, staffing, and rate of commissions.Cadet Command requested participation from the college presidentsof the selected units. Since several presidents declined to participatein the experiment, it was necessary to search for additional schools.As a result, the final list of experimental schools entailed some com-promises. Nonetheless, we believe that the list of schools is repre-sentative of the range of SROTC units, with the exception of militarycolleges and extension centers, which were excluded by design.

Table A.1 reports the schools participating in the contracted former-military and civilian option and their staffing during the experiment.Table A.2 presents a statistical profile of these units in comparisonwith the overall set of SROTC battalions. The test units are wellmatched to the characteristics of the full set of SROTC programs.

Table A.1

Colleges and Universities Participating in the 1997–1998 Staffing Experiment forContracted Former Military and Other Civilians

PMS/APMS Trainer Admin/Log

Academic Institution Replacements Military Replacements Military Replacements Military

Rochester Institute of Technology 2 2 1 1 1 0Providence College 2* 3 1 2 1 0Campbell University 2 4 1 2 1 0University of Arkansas-Pine Bluff 1 2 1 1 0 1Grambling State University 2* 2 1 1 1 1University of Michigan 2 2 1 1 1 0University of Colorado-Colorado Springs 2 2 1 1 0 0Florida State University 2 3 1 1 1 1Truman State University 2 2 1 1 0 0University of Cincinnati 2* 2 1 1 0 0Western Illinois 2 3 1 2 1 1University of Texas Pan American 1 2 1 1 0 0Northwestern State University of Louisiana 2 3 1 1 1 1Pittsburg State University (Kansas) 2* 2 0 2 0 1University of Utah 2* 2 1 1 0 0

TOTAL Second Year 28 36 14 19 8 6 ( First Year 23 41 14 19 8 6 )

Replacement rate (average over both years) 40% 42% 57%

SOURCE: Cadet Command, October 15, 1997.*Indicates one fewer replacement scheduled to be in effect during the first year of the test. In these schools, it was not feasible toimplement all replacements at the start of the test because of Army personnel rotations.NOTE: Desired rate of replacements is 50 percent for PMS/APMS, 50 percent for trainers, and 100 percent for admin/log.

46Staffin

g Arm

y RO

TC

at Co

lleges and

Un

iversities

Additional Details of the Test Plan 47

Table A.2

Comparison of Experimental Units to Overall SROTC Program

MeasureExperimental

UnitsOverall

Program

Number of schools 15 239

ProductionAverage commissions 1992–1996 12.9 13.4

Staffing assignedOfficers 4.2 4.0NCOs 2.9 3.0Civilians 1.4 1.3

Institutional typePublic 80% 72%Private (Scholarship Tier IA) 7% 10%Private (other) 13% 19%

Ethnic marketHistorically black colleges and universities 13% 9%Hispanic-serving institutions 7% 6%

GeographyRegion 1 (East) 27% 35%Region 2 (Midwest) 47% 35%Region 4 (West) 27% 30%

NOTE: Extension centers, military colleges, schools on the Program EfficiencyClosure list, and schools with 1997–1998 reserve option staffing are excludedfrom the overall statistics.

Because of the need to accommodate personnel rotations, severalunits do not have the desired level of replacements, especially duringthe first year of the test. Some of this deficiency will be made upduring the second year. The overall replacement rate for officers(PMS and APMS) is 40 percent averaged over the two years of thetest, compared to a desired 50 percent replacement. For trainers, theoverall replacement rate is 42 percent compared to 50 percent de-sired; for admin/log the overall replacement rate is 57 percent com-pared to 100 percent desired.

We believe that the admin/log replacements will have little effect onthe unit performance, since these functions are less connected tocadet retention and training. The lower level of replacements for


APMS and trainer positions, especially during the first year, will makeit more difficult to detect effects in the experiment, compared to thedesign presented in the body of the report. On the other hand, asTable A.2 shows, the units in the test are fairly well balanced againstthe overall population of units. This balance (which was not as-sumed in the calculations in Chapter Three) will improve the chanceof detecting effects during the experiment. It is unclear whether theadvantage of balancing will offset the difficulty stemming from lowerreplacements.

Plans are now being made to implement the reserve option startingin school year 1998–1999. The VCSA directed that the reserve optionto be tested in the experiment combine AC, AGR, part-time re-servists, and contracted former military personnel to cover instruc-tion and training duties. This mixed alternative is less desirable thana pure-reservist option from the standpoint of empirically assessingthe efficacy of reservist replacement staff. However, it is closer to thetypical staffing that might be found in a given unit if both staff re-placement options are given a green light for implementationthroughout SROTC; that is the reason it was agreed to by the VCSApursuant to a discussion among the general officers attending thetest-approval briefing. As in the former-military option, admin/logpositions would be covered by contracted civilians.

The mixed alternative also differs from our description in terms ofthe composition of the reservist groups used to replace instructorand trainer positions. Unlike the plan described in the main text, thetest will not use reservists for part-time admin/log functions as partof the group replacing instructors and trainers. This change willmost likely increase the resource costs of staffing instructor posi-tions, because up to eight officers will be used, instead of the mix ofofficers and junior enlisted described in the main text. But there isuncertainty over both the grade distribution of officers used in APMSpositions and the extent of NCO use in these positions. Therefore,we do not know the magnitude of the effect on resource costs for thereserve option as tested. Since the trainer positions relied much lesson these part-time admin/log replacements, the impact will besmaller on trainer costs. The test will generate data on the actualgrade distribution employed in each position. This data can guidefuture estimates of resource costs.

Additional Details of the Test Plan 49

In the 1997–1998 school year, three battalions began to implementvariations of these combinations, including TPU reservists. Bycombining reserves with contracted former military, it will be easierto staff the reserve option units. The combination plan is more likelyto achieve the full 15 units required for the test. On the other hand,the test will be harder to interpret. Separating the effects of reservistsfrom those of the former military may be difficult, since the mixedoption will introduce both types of replacement staff into the exper-imental battalions simultaneously. The results of the trial of theformer-military option, which will precede the mixed option by atleast one year, will help to calibrate the measurements in the mixedoption battalions. For example, if the former-military option has un-desirable outcomes and the mixed option does not, the success canlikely be attributed to the reserve personnel. If the former-militaryoption succeeds and the mixed option has undesirable outcomes,the failure can likely be attributed to the reserve personnel. If bothoptions succeed or both fail, it may be difficult to judge how well thereserve personnel would have performed on their own.

Based on this test implementation, the final results of the experimentwith former military personnel and other civilians are projected to beknown in December 1999. If the mixed option can be staffed at 15battalions starting in school year 1998–1999, those results would beknown in December 2000. Interim results will be available in late1998 for the former-military option and late 1999 for the mixedoption.

51

Appendix B

MILITARY PERSONNEL COSTS

We conducted an analysis of personnel costs in order to compare thecost of AC, RC, and civilian personnel. The main text describes thedata and methods used to project the costs of contracted civilians.The analysis of military personnel costs is documented here.

MILITARY COMPENSATION

Military compensation is composed of several pays, allowances, andother benefits. Military pay includes basic pay and a number of spe-cial and incentive pays such as flight pay, hazardous duty pay, andbonuses. Allowances are provided to members to offset the cost ofliving in the form of allowances for quarters, subsistence, and cloth-ing. Other benefits include factors such as the availability of medicalcare, retirement pay, and other similar considerations. However, notall members of the Armed Forces receive the same compensation el-ements nor at the same rate. Many of these costs vary based on theservice member’s grade, longevity, and component of military ser-vice.

We included the following elements of cost in our analysis: BasicPay, Basic Allowance for Quarters (BAQ), Variable Housing Al-lowance (VHA), Basic Allowance for Subsistence (BAS), clothing-replacement allowance, retirement pay accrual, military medicalcare, and FICA. Added to this amount for Active Component per-sonnel was the cost of permanent change of station (PCS). Table B.1lists the AC and RC entitlements to these pays and benefits. Asshown in the table, RC pays and benefits differ during inactive dutytraining (IDT) drills and annual training (AT).


Table B.1

Federal Benefits for Armed Forces Personnel

Benefit

ActiveComponent

Personnel

Reservists DuringInactive Duty Training

(IDT) Assemblies

Reservists DuringAnnual Training

(AT)

Pay andallowances

Basic Pay, BAS,BAQ, VHA, Cloth-ing, and SpecialPay, if authorized

Basic Pay onlycomputed at 1 day payfor each drill period

Basic Pay, BAS,BAQ, and SpecialPay, if authorized

Retirement Upon retirementfrom active-dutyservice

1 retirement point perUTA (maximum 75points per year)

1 retirement pointper day (maximum365/366 points peryear)

Medical care forservice member

Yes Only if duty-related Yes

Medical care forfamily members

Yes No No if on active dutyfor less than 30days

FICA Yes Yes Yes

PCS costs Yes Does not apply Does not apply

SOURCES: 1996 Uniformed Services Almanac and 1996 Reserve Forces Almanac.

COST ELEMENTS

The following describes the assumptions and data used to calculatethe costs for each cost element.

Basic Pay

Under the military pay structure, basic pay by pay grade increases asthe soldier’s time in service increases. This accrual in seniority cansubstantially affect the cost differences between populations, espe-cially if one population is significantly senior to another. Therefore,our cost model estimated the average cost associated with each paygrade based on the distribution of military personnel strengthsacross years of service.

Military Personnel Costs 53

Strength figures by grade and time-in-service for both AC and RCpersonnel were obtained from the Defense Manpower Data Center(DMDC) Information Delivery System. Strength figures were as ofJune 1996. We also used in our analysis the military compensationrates that went into effect January 1, 1996, contained in the 1996 Uni-formed Services Almanac and 1996 Reserve Forces Almanac.

Basic Allowance for Quarters (BAQ)

Active Component members (and reservists during AT) receive anallowance to offset the cost of living in nongovernment quarters.Reservists in an inactive-duty status, such as unit training assemblies(UTAs), are not authorized an allowance for quarters. Basic Al-lowance for Quarters (BAQ) also varies by pay grade and familymember status (i.e., with dependents or without).

We used the BAQ rates that went into effect January 1, 1996, con-tained in the 1996 Uniformed Services Almanac. The family memberstatus of AC and RC personnel was obtained from the DMDC Infor-mation Delivery System. Using these rates and family member sta-tus, we calculated the average amount by grade paid to servicemembers in nongovernment quarters. For the purposes of thisstudy, the (higher) “with dependents rate” was applied to all servicemembers occupying government quarters to establish an “in-kind”rate for their government housing, regardless of their family status.This represents only an estimate of the government’s cost of provid-ing quarters.

Variable Housing Allowance (VHA)

Active Component members also receive a Variable Housing Al-lowance (VHA) to supplement their allowance for quarters in high-housing-cost areas. VHA is only paid to reservists activated for morethan 139 days. As with BAQ, VHA varies by pay grade and familymember status of the service member.

We used the average VHA rate by pay grade from the DoD SelectedMilitary Compensation Tables January 1996 Pay Rates, published bythe DoD Directorate of Compensation. The family member status ofservice members was obtained from the DMDC Information Delivery


System. For the purposes of this study, the “with dependents rate”was applied to service members occupying government quarters toestablish an “in-kind” rate for this allowance.

Basic Allowance for Subsistence (BAS)

Active Component members and reservists during AT receive an al-lowance to offset the cost of meals. Reservists in an inactive-dutystatus, such as unit training assemblies, are not authorized an al-lowance for subsistence (BAS). We used the “authorized to messseparately” rates for officers and enlisted personnel that went intoeffect January 1, 1996, contained in the 1996 Uniformed Services Al-manac.

Clothing-Replacement Allowance

Active Component enlisted members are issued a standard wardrobewhen they enter active duty and receive a clothing replacement al-lowance each year. The replacement allowance is paid at the “basic”rate for members with less than three years of service and a“standard” rate for members with three or more years of service. Ac-tive Component officers receive a one-time payment upon commis-sioning to buy clothing and do not receive any allowance to maintaintheir uniforms. Reservists are not usually eligible for a clothing re-placement allowance unless they are called to active duty for morethan six months. We used the “clothing-replacement allowance”rates for enlisted personnel that went into effect October 1, 1995,contained in the 1996 Uniformed Services Almanac.

Retirement Benefit Accrual

Members of the Armed Forces accrue retirement benefits based ontheir service component and the retirement plan in effect when theyentered military service. Active Component personnel receivemonthly retirement income from the military retirement systemupon retirement from active duty. Reserve Component personnelearn retirement points during inactive-duty training and annualtraining and must achieve 20 satisfactory years of service and be 60years old before retirement pay commences. (Satisfactory years of


service are ones in which the reservist earns a minimum of 50 re-tirement points.)

There are also three distinct nondisability benefit formulas withinthe military retirement system. Military personnel who entered theArmed Services before September 8, 1980, receive retirement paybased on their terminal basic pay. For members entering service onor after September 8, 1980, a “high-three” average of basic pay isused in the computation. In addition, members entering the ArmedServices on or after August 1, 1986, are subject to a penalty if they re-tire with fewer than 30 years of service (until they reach age 62).

The true effect of staffing decisions on retirement costs depends onhow the Army alters its force structure, for example reducing acces-sions and/or reducing various promotion rates. The conventionalbudgeting approach uses retirement accrual normal cost percentage(NCP) rates for each retirement benefit formula to calculate the re-tirement benefit cost. We used current years of service to indicatewhich benefit formula applied, assuming that the service membershad served continuously for those years of service. This calculationused the retirement NCP rates in effect during FY96 contained in theValuation of the Military Retirement System dated September 30,1996, published by the DoD Office of the Actuary.

Military Medical Care

Active Component and Reserve Component personnel during AT areentitled to health care services provided by the military health caresystem. Reservists while on inactive-duty training only have limitedaccess to military medical care. Family members of AC personnel arealso eligible to receive health care from the military health care sys-tem. Reserve family members are not eligible during short-term re-serve duty.

The cost to DoD to provide health care services differs by beneficiarycategory, e.g., for service members as opposed to family members.We used the cost per user by beneficiary category for FY96 to calcu-late the cost of the military health care benefit by pay grade. We val-ued the health care benefit for reservists with limited access duringinactive-duty training at 50 percent of the total cost for full-time ser-vice members.


The cost per user by beneficiary category was provided by the Assis-tant Secretary of Defense for Health Affairs Health Budget and Pro-gram. We used the family member status of service members fromthe DMDC Information Delivery System to determine the number offamily members eligible for military health care services.

FICA

The government pays Federal Insurance Contributions Act (FICA)taxes on all military members’ basic pay. We applied the current rateof 7.65 percent to the basic pay for each pay grade across years ofservice.

Permanent Change of Station (PCS)

Active Component service members and their family members areprovided transportation, movement of their personal property, andreimbursement of relocation expenses when a permanent change ofstation (PCS) is required. We used the average PCS costs for FY96 toestimate the annual PCS cost for officers and enlisted personnel bygrade, averaging over all members. The total PCS costs for FY96 arecontained in the FY97 Budget Estimates published by the Depart-ment of the Army, Military Personnel.

ANNUAL COMPENSATION RATES

We combined all the pay and benefits detailed above into annual av-erages for each grade and applied one-year inflation of 3.2 percent totranslate the 1996 rates into 1997 dollars. Table B.2 displays theseannual compensation rates for the AC and RC by grade and groupedby SROTC position. We calculated the annual compensation for RCpersonnel using a standard 48-drill inactive-duty training year and18.5-day AT. A typical schedule for a reservist in a TPU involvesdrilling one weekend per month during the year (48 drills) plus twoweeks during the summer (14-day AT). Because of the need to staffSROTC summer camps and accompany the battalions on weekendexercises during the school year, we estimate that reservists wouldaverage 48 drills plus 18.5 days AT. The increased costs for the extra


Table B.2

Annual Compensation Rates (1997$)

Position Active Reserve

and Grade Compensation Distribution Compensation Distribution

Instructor

O-6 128,068 0.0% 16,938 7.0%O-5 107,597 0.0% 14,029 24.8%O-4 88,262 28.8% 11,662 29.1%O-3 71,006 71.2% 9,797 39.1%Average 75,981 11,888

Trainer

E-8 66,510 10.8% 8,225 0.0%E-7 57,915 89.2% 7,017 100.0%Average 58,845 7,017

Admin/Log

E-6 49,870 50.9% 5,557 19.3%E-5 41,993 49.1% 4,885 27.2%E-4 34,298 0.0% 3,998 53.5%Average 45,999 4,540

4.5 days AT would represent incremental cash costs for using RC per-sonnel in SROTC.

We assumed that reservist compensation for each grade would bethe same for members of the IRR or Selected Reserves. As noted inthe text, IRR members could function in two different statuses, but atequivalent compensation. They could join the Selected Reserves andparticipate in SROTC through a TPU, earning pay and benefits as de-scribed above. Alternatively, they could remain in the IRR but be ac-tivated with a specified rate of pay for each period worked.(Although more complicated, this method might be employed be-cause these IRR members would not count against the endstrengthceiling for the Selected Reserves as long as they do not serve on con-tinuous duty for more than 180 days.)

To compute average costs for each AC and RC position, we usedgrade distributions. For AC positions to be replaced, we used an es-timate of the grades that would likely be subject to replacement foreach current position. For RC replacements, we used estimates of


the populations that would be available to fill each position. For in-structors, we used the distribution of O-3 through O-6 TPU memberswhose civilian occupations are in secondary or postsecondaryteaching. For trainers, we used the average E-7 pay rate. For theadmin/log support positions, we used the population distribution ofE-4 through E-6 in the OCAR-designated admin/log MOSs.1 Popu-lations were derived from SIDPERS using the appropriate criteria. Tocompute the estimated costs of the TPU/IRR SROTC battalionstaffing, we combined these average rates with the staffing plan inChapter Two of the report.

______________1This admin/log support position is used to fill part of the replacements for APMS andtrainer positions. It is not related to the full-time admin/log staff position.

59

Appendix C

GEOGRAPHIC AVAILABILITY MODEL

With the current AC and AGR staffing for SROTC battalions, the Armyorders soldiers to report to the college locations as required to staffeach battalion. Alternative staffing plans would not allow the Armyto directly order personnel to fill vacancies. Instead, positions wouldbe filled by reservists or contracted civilians.

Since reservists are typically employed (aside from their part-time re-serve duty), they may well be unlikely to move in order to accept apart-time SROTC position. We developed a geographic availabilitymodel to estimate the number of reservists who would be able tostaff SROTC positions, based on the population of reservists within50 miles of each SROTC program.

The contractor employing civilians, including former military, mightbe able to induce potential staff to locate near available positions.But the preferences of individuals—and the limited permitted jobtenure for former military—would reduce the chance that civilianswould locate much differently from current patterns. We thereforeestimated the availability of civilians within 50 miles of each SROTCprogram using a geographic model based on the actual location ofcivilians with the required qualifications as reported in the 1990 U.S.Census.

The data for civilians is derived from the 5 percent Public Use Micro-data Sample (PUMS) of the 1990 U.S. Census, weighted to representthe entire U.S. population. We selected the observations with theappropriate military experience, education, and occupation as de-scribed in Chapter Two of the main text. Separate pools of observa-tions were constructed for instructor, trainer, and admin/log eligible


populations. As described, we dropped observations whose reportedwages exceeded the thresholds for each pool.

We then calculated a scaling percentage for instructors, trainers, andadmin/log by combining all of the factors described in the main text.These factors included military force size reductions and estimatedawareness of and interest in the jobs.

Using the present list of SROTC battalions, we applied the proposedreplacement staffing plan to derive the number of instructors, train-ers, and admin/log positions desired at each school. We termed thisdesired number of positions the demand at each school.

The geographic model then matched the estimated pool of availablepersonnel to the desired replacements. For each school, we used theschool zip code to derive latitude and longitude from a database de-signed for that purpose to calculate the number of potential re-placements within 50 miles of the school.

For the observations from the PUMS, individual-level zip codes arenot provided (to preserve confidentiality). Instead, observations arecoded in Public Use Microdata Areas (PUMAs), which encompassdefined areas. Urban PUMAs are generally small, though ruralPUMAs can be large. For all observations, we allocated the personsin each PUMA to the individual zip codes in the PUMA according tothe general population in each zip code.

A key technical concern was to allocate as many people as possible tothe positions. Each school could draw from an area within 50 miles.Since a number of schools had overlapping areas, we developed amodel to handle the overlapping areas without assigning more of thepool to any school than that school’s demand. The model madeseveral passes through the list of schools to optimally allocate thepool of observations to school positions.

We used the same basic model to estimate the geographic availabilityof TPU and IRR reserve personnel. Since Army databases listedhome zip codes for each reservist, no preliminary allocation was re-quired as with the PUMAs. Otherwise, the model proceeded in thesame way as for the civilians.

Geographic Availability Model 61

Early model runs pointed up that the key constraints on reserve re-placements were the reservists qualified as part-time instructors andtrainers, rather than the part-time admin/log support personnel.Therefore, the final estimates of the reservists’ ability to replace full-time military are based only on the instructors (four TPU/IRR perposition) or trainers (six TPU/IRR per position). In many schools, themodel computed that there would not be sufficient personnel tomeet the full demand. We computed the total number of full-timepositions that could be replaced. We consider one full-time instruc-tor position replaced if the school had more than 0.5 of the requiredTPU/IRR instructors but less than 1.5. If the school had more than1.5 but less than 2.5, we considered two positions replaced, and soforth. The same applied to trainer positions. After computing thenumber of full-time positions replaced at each school, we summedover all schools to compute the total.

63

Appendix D

STATISTICAL POWER CALCULATIONS

The main text indicates the basic methods used to estimate the sta-tistical power of the measures for commissions, workload, and Ad-vanced Camp scores.

NUMBER OF COMMISSIONS

We developed the power calculations for number of commissions us-ing historical data on cadet progress and commissioning. As notedin the main text, the calculations reported are based on changes inthe rate of noncontracted MSIIs (sophomores) contracting as MSIIIs(juniors).

We estimated power using a Monte Carlo simulation. The simula-tion procedure selected the desired number of experimental units atrandom from the pool of eligible units. Each unit was equally likelyto be selected, and no attempts were made to balance the selectedexperimental units with the overall population. To account for asecond simultaneous experimental group, the procedure deleted anequal number of units from the pool. The remainder formed thecomparison group. The simulation used two recent consecutiveyears of data to compute the rate of noncontracted MSIIs in the firstyear contracting as MSIIIs in the second year of the data for eachschool.

In the comparison group, the simulation used the raw historical data.In the experimental group, the historical data were adjusted to simu-late the effect of a performance degradation. Specifically, the simu-lation multiplied a disturbance by the historical data. Because we


believe that the alternative staffing might change both the mean andvariance of the measure, the disturbance we used was a normallydistributed pseudorandom variable with mean set at the specifiedeffect size (e.g., 15 percent). The standard deviation of the distur-bance (at the unit level) was equal to the effect size divided by thesquare root of the raw number of observations that contributed tothe rate (i.e., the number of noncontracted MSIIs in the first-yeardata). The standard deviation is thus larger in small units andsmaller in large units, consistent with averaging individual cadet ob-servations into a unit average. The introduction of this standard de-viation component is more conservative than many approaches toexperimental design, which would only change the mean in the ex-perimental group.

The simulation then ran a logit on the rate with a dummy variableindicating the experimental group. If the estimated coefficient onthe dummy variable was significant at the two-tailed 0.10 level, thattrial was scored as a success. If the coefficient was not significant,that trial was scored as a failure. The procedure then drew a newrandomly selected experimental group and repeated. We reportedthe percentage of 1,000 trials that were successful for each experi-mental group size (5, 10, 15, etc.) and for each effect size (5 percent,15 percent, and 25 percent).

We ran some simulations including a school’s prior rate of contract-ing (from earlier data) as a control in the logit estimation, but suchcontrols yielded very small improvements, due to the wide variationfrom year to year. (The year-on-year correlation of the rates wasabout 0.1.)

WORKLOAD

Unlike the commissions and camp scores, there are no detailed his-torical data to use in estimating statistical power for workload.

Based on our judgment and some field observations during the pilottest of the workload survey, we assumed that each observation of aspecific individual’s workload would have a standard deviation of 10hours per week, or a variance of 100. The variance was assumed toconsist of five independent components: school-specific (12.5),

Statistical Power Calculations 65

school-year (12.5), person-specific (25), week-specific (25), and ran-dom variation (25).

The unit of analysis here was the average reported workload over allfull-time military respondents during the survey period. The compu-tations were based on the variance that resulted from seven adminis-trations of a workload survey per year. On average, each experimen-tal battalion would have three full-time military cadre members re-sponding to the survey during each year. Because military personnelrotate on a three-year schedule, we would expect one of the three torotate and be replaced in the second year of a two-year test. Thus,there would be four different individuals participating in a two-yeartest.

In a one-year test, each experimental school contributes one obser-vation on the school-specific and school-year component, three ob-servations on the person-specific, seven on the week-specific, andtwenty-one (three persons times seven weeks) on the random varia-tion. The variance of the average is thus 12.5 + 12.5 + 25/3 + 25/7 +25/21 = 38.1.

In a two-year test, each experimental school contributes one obser-vation on the school-specific component, two on the school-year,four observations on the person-specific, fourteen on the week-specific, and forty-two (three persons times seven weeks per yeartimes two years) on the random variation. The variance of the aver-age is thus 12.5 + 12.5/2 + 25/4 + 25/14 + 25/42 = 27.4.

For simplicity, we assumed that the variance in the comparisongroup was the same as in the experimental group. (Since the com-parison schools are not replacing military personnel, those schoolshave about twice the number of full-time military during the experi-ment compared to the experimental schools. The variance of aver-age workload in the comparison schools might therefore be lowerthan in the experimental schools because of the greater number ofindividuals contributing to the average in the comparison schools.But the differences in structure between experimental and compari-son schools could introduce other sources of variation.)

The power calculations were computed using the formula for com-parison between two groups with known variance (implemented inthe Stata sampsi procedure). Although the variance would not be


known exactly, the large size of the comparison group means that thevariance could be estimated very precisely. In this light, the simplifi-cation of using the formula for known variance comparisons seemedreasonable.

ADVANCED CAMP SCORES

Power calculations for Advanced Camp scores were based on aMonte Carlo procedure similar to that used for the number of com-missions. Because the historical data report individual scores atcamp, we used the individual-level data.

As in the earlier procedure, the Monte Carlo simulation randomlyselected units for the experimental group and comparison group.The simulation ran ordinary least squares regression on the campscores, with a dummy variable indicating if the cadet record be-longed to a school designated as an experimental unit.

Because of the potential for common factors to affect the perfor-mance of cadets from the same school, we adjusted the resultingnumber of units required to achieve a given level of power by a com-puted design effect.

In this situation, the design effect is given by the F-statistic from aone-way ANOVA of camp score on school. For this data, the F-valuewas 1.7. To be slightly conservative, we set the design effect a bithigher, at 2.

The reported power percentages are based on 1,000 trials for eachgroup size and effect level.

67

REFERENCES

Coumbe, Arthur T., and Lee S. Harford, U.S. Army Cadet Command:The 10 Year History (Cadet Command Historical Study Series),Washington, D.C.: U.S. Government Printing Office, 1996.

Defense Manpower Data Center (DMDC), DMDC Information Deliv-ery System Version 2.0, Monterey, California, 1996.

1996 Reserve Forces Almanac, Falls Church, VA: Uniformed ServicesAlmanac, Inc., 1996.

1996 Uniformed Services Almanac, Falls Church, VA: UniformedServices Almanac, Inc., 1996.

SAS Institute, SAS Release 6, Cary, North Carolina, 1997.

Stata Corporation, Stata Release 4, College Station, Texas, 1996.

U.S. Army Military Personnel Center, Standard Installation/DivisionPersonnel System (SIDPERS), Alexandria, Virginia, 1996.

U.S. Bureau of the Census, Current Population Survey, Washington,D.C., various editions.

U.S. Bureau of the Census, Public Use Microdata Sample, Washing-ton, D.C., 1990.

U.S. Department of the Army, Headquarters Department of the ArmyDecision Support System (HQDADSS), Washington, D.C., 1996.

Wolfram Research, Mathematica Version 3.0, Champaign, Illinois,1997.

staffing army rotc at colleges and universities · ... the other employs drilling reservists ......

Documents